Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Mobileye will launch a self-driving taxi service in Israel in 2019

This site may earn commission on affiliate links.
Plus theres no diff between eyeq4 and eyeq5 other than more tflops and being an open system for other manufacturers.

Er, no. There's definitely a diff.

Source: The Evolution of EyeQ - Mobileye

mobileye_eyeq.png
 
The difference IS the open system i referred to, vision central computer = eyeq4 software. open software platform is the only new thing. With an open platform comes the need for hardware security.
Something that has not been completely clear to me regarding Intel/Mobileye eyeq series has been what sort of interface they provide.

Would it be correct to say than from eyeq1-4 it was a blackbox for the oem in a sense?
In other words what is the typical api that the eyeq4 chip provides?
Is it a list of lane line positions, pedestrian pixel/coordinates, and a few other simple values?
Is it more elaborate than this?
For example, do I feed the eyeq chip data from the radar??

Ie. How does GM uses the same chip as few other automakers on their cars (nissan, lexus, toyota?) and they have different capabilities?
Is that something that GM coded by reading the data from the eyeq chips?
Do you have no choice but to pay extra to mobileye to provide extra features??
 
open software platform is the only new thing.

No, the open vs closed is not the only new thing - they clearly expect a continuing evolution in functionality in terms of levels of autonomy (4-5 from 3) and in terms of driver interaction (Mind Off vs Eyes Off).

What documents are you reading?

Feel free to provide EVIDENCE from Mobileye which contradicts this:

mobileye_eyeq.png
 
There is a YouTube video of the mobileye system in action. It seemed ahead of what Tesla has deployed already however journalist commented it was scary at times therefore the video probably cut out the bad parts.
 
Latest news is (as of today) from May 2018: Mobileye In The News - Mobileye

Which is a kind of Mad Max demo around Jerusalem: A lap of Jerusalem in Intel's surprisingly aggressive self-driving car

together with speculation/promise of future work.

Mad max? Please dont compare mobileye system to that wack sh*t.

No, the open vs closed is not the only new thing - they clearly expect a continuing evolution in functionality in terms of levels of autonomy (4-5 from 3) and in terms of driver interaction (Mind Off vs Eyes Off).

What documents are you reading?

Feel free to provide EVIDENCE from Mobileye which contradicts this:

No that is simply marketing. A eyeq3 can be used for L3 sdc for example. Proven by the Audi A8 using a eyeq3. Just as eyeq4 can be used for level 4 and 5 sdc proven by the fact several companies today are using it for their level4 and 5 systems.

Aptiv (Delphi) uses eyeq4.

Notice that there is no actual new feature or listed feature improvements listed under eyeq5.
 
Last edited:
Aren’t Nissan also “innovating” with announcements about the same thing in Japan?

Nissan systems uses mobileye eyeq4 chips.

"Disappoint" might be a better way of putting it.



This is just a prototype fleet, will probably have a massive server installed in the car, loaded with buggy dev software and remote-piloted through anything more complex than a fork in the road. They are years behind Waymo.

Jaguar and Waymo are teaming up to make 20,000 self-driving i-Paces

They said it wont be a prototype fleet and cars have massive servers because it's easier to debug each individual systems that way during development. Nevertheless the car will use 4x eyeq4.

Lastly google cars struggle to merge during dense traffic because of how conservative they are programmed. Mobileye driving policy handles this flawlessly in a more aggressive Isreal.
Now is waymo more validated than mobileye driving policy? Ofcourse. But for how long though? Because what mobileye is gunning for is a system they can drop anywhere in the world and all you need to do is adjust a few preferences and bam, its fully self driving. Waymo? Not so much.


Wait...

And the project will start in early 2019 with a few dozen vehicles and scale to “hundreds of self-driving electric cars” by 2022.

Only hundreds of cars in 2022? Total? Globally? Or is Mobileye planning to launch in multiple cities, and/or supply the Mobileye AV kit to OEMs? If it’s literally just a few hundred cars in one city in 2022 — and that’s all Mobileye is planning to do — Mobileye is setting its sights surprisingly low. I was hoping it would be aiming to supply AV technology to millions of cars by then.

In 2019, will the cars have safety drivers? This wasn’t mentioned in the article. It seems only responsible to have safety drivers at least for the first 65 million miles, or whatever number is required to statistically demonstrate that the crash rate for the autonomous vehicles is lower than the rate for human drivers. So, I hope there will safety drivers initially.

Same for Waymo, unless they have already done enough miles to have good statistical evidence.

They plan to expand to other cities and sell the AV kit to other companies. This partnership is quite different from the bmw, aptiv or Nissan. Because its fully mobileye written software end to end.

Their plan is to get it self driving in isreal and then drop it anywhere in the world and have it work there too. That would be a true av kit system and they are making it happen because of REM mapping in prod uploading map data today.
 
Last edited:
I think it's just as exciting or more exciting than Waymo because it can demonstrate that the technology isn't limited to fleet vehicles.

Sure it's a taxi/fleet type thing, but it's just to demonstrate it. The important thing is Intel/Mobile eye is a solution provider unlike WayMo/Cruise which might force the technology to only being for fleet cars.

As someone who wants self-driving in a car I own I say this is exciting news.

The only thing not very exciting is it's in Israel so I can't exactly try it out. Can't wait for a self-driving taxi service to arrive in Seattle.

I don't think we want the autonomous section of this website to exclude anything/anyone that's related to autonomous driving. MobileEye is absolutely one of the most important companies when it comes to self-driving cars. There is a reason Intel bought it for 15 billion dollars.
 
No that is simply marketing. A eyeq3 can be used for L3 sdc for example. Proven by the Audi A8 using a eyeq3. Just as eyeq4 can be used for level 4 and 5 sdc proven by the fact several companies today are using it for their level4 and 5 systems.

Nice! My main motivation for buying a Tesla was the Autopilot. I make 95% of my daily commute on Autopilot.

I want that level 4 or 5 you talk about, and I would be willing to change my Tesla for an alternative with this superior level of autonomy.

Could you tell me which brand / model should I buy now, today, on 2018 for that level 4/5? Please note that I live on a small town near Madrid (Spain/Europe). Half of my daily commute are on secondary roads, where we have just one lane on each direction and the paintings are not very clear (although Autopilot interprets and drives them as good as I do).

Please, no demos, videos, dreams nor future availability, for that we already have the 2016 Tesla FSD demo which at this point we all agree it's just bullshit.
 
Could you tell me which brand / model should I buy now, today, on 2018 for that level 4/5?
tenor.gif


Please note that I live on a small town near Madrid (Spain/Europe). Half of my daily commute are on secondary roads, where we have just one lane on each direction and the paintings are not very clear (although Autopilot interprets and drives them as good as I do). Please, no demos, videos, dreams nor future availability, for that we already have the 2016 Tesla FSD demo which at this point we all agree it's just bullshit.
tenor.gif
 
Mobileye - Generating Augmented Corner-Case Lane Markings for NN training

Detecting lanes and road markings are some of the great challenges autonomous driving technologies face today. In Mobileye, we do it by training neural networks. This method requires an enormous amount of data from many countries, different weather conditions etc. However, in some road scenarios the available data set is still too small, or the data is not accurate. In this session, I will discuss how we developed a system that uses Python with OpenCV to generate augmented images and matching ground truth data in order to improve accuracy of neural networks.

 
Last edited:
Even in the mobileye presentation lines are not very realistic (other than perspective corrected and blended).

Also, I do not see how texture or scenery would or should matter to an autonomous system therefore road with all the markings could be fully computer generated.